includes/normal/UtfNormal.php

   1 <?php
   2 # Copyright (C) 2004 Brion Vibber <brion@pobox.com>
   3 # http://www.mediawiki.org/
   4 #
   5 # This program is free software; you can redistribute it and/or modify
   6 # it under the terms of the GNU General Public License as published by
   7 # the Free Software Foundation; either version 2 of the License, or
   8 # (at your option) any later version.
   9 #
  10 # This program is distributed in the hope that it will be useful,
  11 # but WITHOUT ANY WARRANTY; without even the implied warranty of
  12 # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
  13 # GNU General Public License for more details.
  14 #
  15 # You should have received a copy of the GNU General Public License along
  16 # with this program; if not, write to the Free Software Foundation, Inc.,
  17 # 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
  18 # http://www.gnu.org/copyleft/gpl.html
  19
  20 /**
  21  * Unicode normalization routines for working with UTF-8 strings.
  22  * Currently assumes that input strings are valid UTF-8!
  23  *
  24  * Not as fast as I'd like, but should be usable for most purposes.
  25  * UtfNormal::toNFC() will bail early if given ASCII text or text
  26  * it can quickly deterimine is already normalized.
  27  *
  28  * All functions can be called static.
  29  *
  30  * See description of forms at http://www.unicode.org/reports/tr15/
  31  *
  32  * @addtogroup UtfNormal
  33  */
  34
  35 /** */
  36 require_once dirname(__FILE__).'/UtfNormalUtil.php';
  37
  38 global $utfCombiningClass, $utfCanonicalComp, $utfCanonicalDecomp;
  39 $utfCombiningClass = NULL;
  40 $utfCanonicalComp = NULL;
  41 $utfCanonicalDecomp = NULL;
  42
  43 # Load compatibility decompositions on demand if they are needed.
  44 global $utfCompatibilityDecomp;
  45 $utfCompatibilityDecomp = NULL;
  46
  47 define( 'UNICODE_HANGUL_FIRST', 0xac00 );
  48 define( 'UNICODE_HANGUL_LAST',  0xd7a3 );
  49
  50 define( 'UNICODE_HANGUL_LBASE', 0x1100 );
  51 define( 'UNICODE_HANGUL_VBASE', 0x1161 );
  52 define( 'UNICODE_HANGUL_TBASE', 0x11a7 );
  53
  54 define( 'UNICODE_HANGUL_LCOUNT', 19 );
  55 define( 'UNICODE_HANGUL_VCOUNT', 21 );
  56 define( 'UNICODE_HANGUL_TCOUNT', 28 );
  57 define( 'UNICODE_HANGUL_NCOUNT', UNICODE_HANGUL_VCOUNT * UNICODE_HANGUL_TCOUNT );
  58
  59 define( 'UNICODE_HANGUL_LEND', UNICODE_HANGUL_LBASE + UNICODE_HANGUL_LCOUNT - 1 );
  60 define( 'UNICODE_HANGUL_VEND', UNICODE_HANGUL_VBASE + UNICODE_HANGUL_VCOUNT - 1 );
  61 define( 'UNICODE_HANGUL_TEND', UNICODE_HANGUL_TBASE + UNICODE_HANGUL_TCOUNT - 1 );
  62
  63 define( 'UNICODE_SURROGATE_FIRST', 0xd800 );
  64 define( 'UNICODE_SURROGATE_LAST', 0xdfff );
  65 define( 'UNICODE_MAX', 0x10ffff );
  66 define( 'UNICODE_REPLACEMENT', 0xfffd );
  67
  68
  69 define( 'UTF8_HANGUL_FIRST', "\xea\xb0\x80" /*codepointToUtf8( UNICODE_HANGUL_FIRST )*/ );
  70 define( 'UTF8_HANGUL_LAST', "\xed\x9e\xa3" /*codepointToUtf8( UNICODE_HANGUL_LAST )*/ );
  71
  72 define( 'UTF8_HANGUL_LBASE', "\xe1\x84\x80" /*codepointToUtf8( UNICODE_HANGUL_LBASE )*/ );
  73 define( 'UTF8_HANGUL_VBASE', "\xe1\x85\xa1" /*codepointToUtf8( UNICODE_HANGUL_VBASE )*/ );
  74 define( 'UTF8_HANGUL_TBASE', "\xe1\x86\xa7" /*codepointToUtf8( UNICODE_HANGUL_TBASE )*/ );
  75
  76 define( 'UTF8_HANGUL_LEND', "\xe1\x84\x92" /*codepointToUtf8( UNICODE_HANGUL_LEND )*/ );
  77 define( 'UTF8_HANGUL_VEND', "\xe1\x85\xb5" /*codepointToUtf8( UNICODE_HANGUL_VEND )*/ );
  78 define( 'UTF8_HANGUL_TEND', "\xe1\x87\x82" /*codepointToUtf8( UNICODE_HANGUL_TEND )*/ );
  79
  80 define( 'UTF8_SURROGATE_FIRST', "\xed\xa0\x80" /*codepointToUtf8( UNICODE_SURROGATE_FIRST )*/ );
  81 define( 'UTF8_SURROGATE_LAST', "\xed\xbf\xbf" /*codepointToUtf8( UNICODE_SURROGATE_LAST )*/ );
  82 define( 'UTF8_MAX', "\xf4\x8f\xbf\xbf" /*codepointToUtf8( UNICODE_MAX )*/ );
  83 define( 'UTF8_REPLACEMENT', "\xef\xbf\xbd" /*codepointToUtf8( UNICODE_REPLACEMENT )*/ );
  84 #define( 'UTF8_REPLACEMENT', '!' );
  85
  86 define( 'UTF8_OVERLONG_A', "\xc1\xbf" );
  87 define( 'UTF8_OVERLONG_B', "\xe0\x9f\xbf" );
  88 define( 'UTF8_OVERLONG_C', "\xf0\x8f\xbf\xbf" );
  89
  90 # These two ranges are illegal
  91 define( 'UTF8_FDD0', "\xef\xb7\x90" /*codepointToUtf8( 0xfdd0 )*/ );
  92 define( 'UTF8_FDEF', "\xef\xb7\xaf" /*codepointToUtf8( 0xfdef )*/ );
  93 define( 'UTF8_FFFE', "\xef\xbf\xbe" /*codepointToUtf8( 0xfffe )*/ );
  94 define( 'UTF8_FFFF', "\xef\xbf\xbf" /*codepointToUtf8( 0xffff )*/ );
  95
  96 define( 'UTF8_HEAD', false );
  97 define( 'UTF8_TAIL', true );
  98
  99
 100 /**
 101  * For using the ICU wrapper
 102  */
 103 define( 'UNORM_NONE', 1 );
 104 define( 'UNORM_NFD',  2 );
 105 define( 'UNORM_NFKD', 3 );
 106 define( 'UNORM_NFC',  4 );
 107 define( 'UNORM_DEFAULT', UNORM_NFC );
 108 define( 'UNORM_NFKC', 5 );
 109 define( 'UNORM_FCD',  6 );
 110
 111 define( 'NORMALIZE_ICU', function_exists( 'utf8_normalize' ) );
 112
 113 /**
 114  *
 115  */
 116 class UtfNormal {
 117         /**
 118          * The ultimate convenience function! Clean up invalid UTF-8 sequences,
 119          * and convert to normal form C, canonical composition.
 120          *
 121          * Fast return for pure ASCII strings; some lesser optimizations for
 122          * strings containing only known-good characters. Not as fast as toNFC().
 123          *
 124          * @param string $string a UTF-8 string
 125          * @return string a clean, shiny, normalized UTF-8 string
 126      * @static
 127          */
 128         static function cleanUp( $string ) {
 129                 if( NORMALIZE_ICU ) {
 130                         # We exclude a few chars that ICU would not.
 131                         $string = preg_replace(
 132                                 '/[\x00-\x08\x0b\x0c\x0e-\x1f]/',
 133                                 UTF8_REPLACEMENT,
 134                                 $string );
 135                         $string = str_replace( UTF8_FFFE, UTF8_REPLACEMENT, $string );
 136                         $string = str_replace( UTF8_FFFF, UTF8_REPLACEMENT, $string );
 137
 138                         # UnicodeString constructor fails if the string ends with a
 139                         # head byte. Add a junk char at the end, we'll strip it off.
 140                         return rtrim( utf8_normalize( $string . "\x01", UNORM_NFC ), "\x01" );
 141                 } elseif( UtfNormal::quickIsNFCVerify( $string ) ) {
 142                         # Side effect -- $string has had UTF-8 errors cleaned up.
 143                         return $string;
 144                 } else {
 145                         return UtfNormal::NFC( $string );
 146                 }
 147         }
 148
 149         /**
 150          * Convert a UTF-8 string to normal form C, canonical composition.
 151          * Fast return for pure ASCII strings; some lesser optimizations for
 152          * strings containing only known-good characters.
 153          *
 154          * @param string $string a valid UTF-8 string. Input is not validated.
 155          * @return string a UTF-8 string in normal form C
 156          * @static
 157          */
 158         static function toNFC( $string ) {
 159                 if( NORMALIZE_ICU )
 160                         return utf8_normalize( $string, UNORM_NFC );
 161                 elseif( UtfNormal::quickIsNFC( $string ) )
 162                         return $string;
 163                 else
 164                         return UtfNormal::NFC( $string );
 165         }
 166
 167         /**
 168          * Convert a UTF-8 string to normal form D, canonical decomposition.
 169          * Fast return for pure ASCII strings.
 170          *
 171          * @param string $string a valid UTF-8 string. Input is not validated.
 172          * @return string a UTF-8 string in normal form D
 173          * @static
 174          */
 175         static function toNFD( $string ) {
 176                 if( NORMALIZE_ICU )
 177                         return utf8_normalize( $string, UNORM_NFD );
 178                 elseif( preg_match( '/[\x80-\xff]/', $string ) )
 179                         return UtfNormal::NFD( $string );
 180                 else
 181                         return $string;
 182         }
 183
 184         /**
 185          * Convert a UTF-8 string to normal form KC, compatibility composition.
 186          * This may cause irreversible information loss, use judiciously.
 187          * Fast return for pure ASCII strings.
 188          *
 189          * @param string $string a valid UTF-8 string. Input is not validated.
 190          * @return string a UTF-8 string in normal form KC
 191          * @static
 192          */
 193         static function toNFKC( $string ) {
 194                 if( NORMALIZE_ICU )
 195                         return utf8_normalize( $string, UNORM_NFKC );
 196                 elseif( preg_match( '/[\x80-\xff]/', $string ) )
 197                         return UtfNormal::NFKC( $string );
 198                 else
 199                         return $string;
 200         }
 201
 202         /**
 203          * Convert a UTF-8 string to normal form KD, compatibility decomposition.
 204          * This may cause irreversible information loss, use judiciously.
 205          * Fast return for pure ASCII strings.
 206          *
 207          * @param string $string a valid UTF-8 string. Input is not validated.
 208          * @return string a UTF-8 string in normal form KD
 209          * @static
 210          */
 211         static function toNFKD( $string ) {
 212                 if( NORMALIZE_ICU )
 213                         return utf8_normalize( $string, UNORM_NFKD );
 214                 elseif( preg_match( '/[\x80-\xff]/', $string ) )
 215                         return UtfNormal::NFKD( $string );
 216                 else
 217                         return $string;
 218         }
 219
 220         /**
 221          * Load the basic composition data if necessary
 222          * @private
 223          * @static
 224          */
 225         static function loadData() {
 226                 global $utfCombiningClass;
 227                 if( !isset( $utfCombiningClass ) ) {
 228                         require_once( 'UtfNormalData.inc' );
 229                 }
 230         }
 231
 232         /**
 233          * Returns true if the string is _definitely_ in NFC.
 234          * Returns false if not or uncertain.
 235          * @param string $string a valid UTF-8 string. Input is not validated.
 236          * @return bool
 237          * @static
 238          */
 239         static function quickIsNFC( $string ) {
 240                 # ASCII is always valid NFC!
 241                 # If it's pure ASCII, let it through.
 242                 if( !preg_match( '/[\x80-\xff]/', $string ) ) return true;
 243
 244                 UtfNormal::loadData();
 245                 global $utfCheckNFC, $utfCombiningClass;
 246                 $len = strlen( $string );
 247                 for( $i = 0; $i < $len; $i++ ) {
 248                         $c = $string{$i};
 249                         $n = ord( $c );
 250                         if( $n < 0x80 ) {
 251                                 continue;
 252                         } elseif( $n >= 0xf0 ) {
 253                                 $c = substr( $string, $i, 4 );
 254                                 $i += 3;
 255                         } elseif( $n >= 0xe0 ) {
 256                                 $c = substr( $string, $i, 3 );
 257                                 $i += 2;
 258                         } elseif( $n >= 0xc0 ) {
 259                                 $c = substr( $string, $i, 2 );
 260                                 $i++;
 261                         }
 262                         if( isset( $utfCheckNFC[$c] ) ) {
 263                                 # If it's NO or MAYBE, bail and do the slow check.
 264                                 return false;
 265                         }
 266                         if( isset( $utfCombiningClass[$c] ) ) {
 267                                 # Combining character? We might have to do sorting, at least.
 268                                 return false;
 269                         }
 270                 }
 271                 return true;
 272         }
 273
 274         /**
 275          * Returns true if the string is _definitely_ in NFC.
 276          * Returns false if not or uncertain.
 277          * @param string $string a UTF-8 string, altered on output to be valid UTF-8 safe for XML.
 278          * @static
 279          */
 280         static function quickIsNFCVerify( &$string ) {
 281                 # Screen out some characters that eg won't be allowed in XML
 282                 $string = preg_replace( '/[\x00-\x08\x0b\x0c\x0e-\x1f]/', UTF8_REPLACEMENT, $string );
 283
 284                 # ASCII is always valid NFC!
 285                 # If we're only ever given plain ASCII, we can avoid the overhead
 286                 # of initializing the decomposition tables by skipping out early.
 287                 if( !preg_match( '/[\x80-\xff]/', $string ) ) return true;
 288
 289                 static $checkit = null, $tailBytes = null, $utfCheckOrCombining = null;
 290                 if( !isset( $checkit ) ) {
 291                         # Load/build some scary lookup tables...
 292                         UtfNormal::loadData();
 293                         global $utfCheckNFC, $utfCombiningClass;
 294
 295                         $utfCheckOrCombining = array_merge( $utfCheckNFC, $utfCombiningClass );
 296
 297                         # Head bytes for sequences which we should do further validity checks
 298                         $checkit = array_flip( array_map( 'chr',
 299                                         array( 0xc0, 0xc1, 0xe0, 0xed, 0xef,
 300                                                    0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7,
 301                                                    0xf8, 0xf9, 0xfa, 0xfb, 0xfc, 0xfd, 0xfe, 0xff ) ) );
 302
 303                         # Each UTF-8 head byte is followed by a certain
 304                         # number of tail bytes.
 305                         $tailBytes = array();
 306                         for( $n = 0; $n < 256; $n++ ) {
 307                                 if( $n < 0xc0 ) {
 308                                         $remaining = 0;
 309                                 } elseif( $n < 0xe0 ) {
 310                                         $remaining = 1;
 311                                 } elseif( $n < 0xf0 ) {
 312                                         $remaining = 2;
 313                                 } elseif( $n < 0xf8 ) {
 314                                         $remaining = 3;
 315                                 } elseif( $n < 0xfc ) {
 316                                         $remaining = 4;
 317                                 } elseif( $n < 0xfe ) {
 318                                         $remaining = 5;
 319                                 } else {
 320                                         $remaining = 0;
 321                                 }
 322                                 $tailBytes[chr($n)] = $remaining;
 323                         }
 324                 }
 325
 326                 # Chop the text into pure-ASCII and non-ASCII areas;
 327                 # large ASCII parts can be handled much more quickly.
 328                 # Don't chop up Unicode areas for punctuation, though,
 329                 # that wastes energy.
 330                 $matches = array();
 331                 preg_match_all(
 332                         '/([\x00-\x7f]+|[\x80-\xff][\x00-\x40\x5b-\x5f\x7b-\xff]*)/',
 333                         $string, $matches );
 334
 335                 $looksNormal = true;
 336                 $base = 0;
 337                 $replace = array();
 338                 foreach( $matches[1] as $str ) {
 339                         $chunk = strlen( $str );
 340
 341                         if( $str{0} < "\x80" ) {
 342                                 # ASCII chunk: guaranteed to be valid UTF-8
 343                                 # and in normal form C, so skip over it.
 344                                 $base += $chunk;
 345                                 continue;
 346                         }
 347
 348                         # We'll have to examine the chunk byte by byte to ensure
 349                         # that it consists of valid UTF-8 sequences, and to see
 350                         # if any of them might not be normalized.
 351                         #
 352                         # Since PHP is not the fastest language on earth, some of
 353                         # this code is a little ugly with inner loop optimizations.
 354
 355                         $head = '';
 356                         $len = $chunk + 1; # Counting down is faster. I'm *so* sorry.
 357
 358                         for( $i = -1; --$len; ) {
 359                                 if( $remaining = $tailBytes[$c = $str{++$i}] ) {
 360                                         # UTF-8 head byte!
 361                                         $sequence = $head = $c;
 362                                         do {
 363                                                 # Look for the defined number of tail bytes...
 364                                                 if( --$len && ( $c = $str{++$i} ) >= "\x80" && $c < "\xc0" ) {
 365                                                         # Legal tail bytes are nice.
 366                                                         $sequence .= $c;
 367                                                 } else {
 368                                                         if( 0 == $len ) {
 369                                                                 # Premature end of string!
 370                                                                 # Drop a replacement character into output to
 371                                                                 # represent the invalid UTF-8 sequence.
 372                                                                 $replace[] = array( UTF8_REPLACEMENT,
 373                                                                                                         $base + $i + 1 - strlen( $sequence ),
 374                                                                                                         strlen( $sequence ) );
 375                                                                 break 2;
 376                                                         } else {
 377                                                                 # Illegal tail byte; abandon the sequence.
 378                                                                 $replace[] = array( UTF8_REPLACEMENT,
 379                                                                                                         $base + $i - strlen( $sequence ),
 380                                                                                                         strlen( $sequence ) );
 381                                                                 # Back up and reprocess this byte; it may itself
 382                                                                 # be a legal ASCII or UTF-8 sequence head.
 383                                                                 --$i;
 384                                                                 ++$len;
 385                                                                 continue 2;
 386                                                         }
 387                                                 }
 388                                         } while( --$remaining );
 389
 390                                         if( isset( $checkit[$head] ) ) {
 391                                                 # Do some more detailed validity checks, for
 392                                                 # invalid characters and illegal sequences.
 393                                                 if( $head == "\xed" ) {
 394                                                         # 0xed is relatively frequent in Korean, which
 395                                                         # abuts the surrogate area, so we're doing
 396                                                         # this check separately to speed things up.
 397
 398                                                         if( $sequence >= UTF8_SURROGATE_FIRST ) {
 399                                                                 # Surrogates are legal only in UTF-16 code.
 400                                                                 # They are totally forbidden here in UTF-8
 401                                                                 # utopia.
 402                                                                 $replace[] = array( UTF8_REPLACEMENT,
 403                                                                              $base + $i + 1 - strlen( $sequence ),
 404                                                                              strlen( $sequence ) );
 405                                                                 $head = '';
 406                                                                 continue;
 407                                                         }
 408                                                 } else {
 409                                                         # Slower, but rarer checks...
 410                                                         $n = ord( $head );
 411                                                         if(
 412                                                                 # "Overlong sequences" are those that are syntactically
 413                                                                 # correct but use more UTF-8 bytes than are necessary to
 414                                                                 # encode a character. Naïve string comparisons can be
 415                                                                 # tricked into failing to see a match for an ASCII
 416                                                                 # character, for instance, which can be a security hole
 417                                                                 # if blacklist checks are being used.
 418                                                                ($n  < 0xc2 && $sequence <= UTF8_OVERLONG_A)
 419                                                                 || ($n == 0xe0 && $sequence <= UTF8_OVERLONG_B)
 420                                                                 || ($n == 0xf0 && $sequence <= UTF8_OVERLONG_C)
 421
 422                                                                 # U+FFFE and U+FFFF are explicitly forbidden in Unicode.
 423                                                                 || ($n == 0xef &&
 424                                                                            ($sequence == UTF8_FFFE)
 425                                                                         || ($sequence == UTF8_FFFF) )
 426
 427                                                                 # Unicode has been limited to 21 bits; longer
 428                                                                 # sequences are not allowed.
 429                                                                 || ($n >= 0xf0 && $sequence > UTF8_MAX) ) {
 430
 431                                                                 $replace[] = array( UTF8_REPLACEMENT,
 432                                                                                     $base + $i + 1 - strlen( $sequence ),
 433                                                                                     strlen( $sequence ) );
 434                                                                 $head = '';
 435                                                                 continue;
 436                                                         }
 437                                                 }
 438                                         }
 439
 440                                         if( isset( $utfCheckOrCombining[$sequence] ) ) {
 441                                                 # If it's NO or MAYBE, we'll have to rip
 442                                                 # the string apart and put it back together.
 443                                                 # That's going to be mighty slow.
 444                                                 $looksNormal = false;
 445                                         }
 446
 447                                         # The sequence is legal!
 448                                         $head = '';
 449                                 } elseif( $c < "\x80" ) {
 450                                         # ASCII byte.
 451                                         $head = '';
 452                                 } elseif( $c < "\xc0" ) {
 453                                         # Illegal tail bytes
 454                                         if( $head == '' ) {
 455                                                 # Out of the blue!
 456                                                 $replace[] = array( UTF8_REPLACEMENT, $base + $i, 1 );
 457                                         } else {
 458                                                 # Don't add if we're continuing a broken sequence;
 459                                                 # we already put a replacement character when we looked
 460                                                 # at the broken sequence.
 461                                                 $replace[] = array( '', $base + $i, 1 );
 462                                         }
 463                                 } else {
 464                                         # Miscellaneous freaks.
 465                                         $replace[] = array( UTF8_REPLACEMENT, $base + $i, 1 );
 466                                         $head = '';
 467                                 }
 468                         }
 469                         $base += $chunk;
 470                 }
 471                 if( count( $replace ) ) {
 472                         # There were illegal UTF-8 sequences we need to fix up.
 473                         $out = '';
 474                         $last = 0;
 475                         foreach( $replace as $rep ) {
 476                                 list( $replacement, $start, $length ) = $rep;
 477                                 if( $last < $start ) {
 478                                         $out .= substr( $string, $last, $start - $last );
 479                                 }
 480                                 $out .= $replacement;
 481                                 $last = $start + $length;
 482                         }
 483                         if( $last < strlen( $string ) ) {
 484                                 $out .= substr( $string, $last );
 485                         }
 486                         $string = $out;
 487                 }
 488                 return $looksNormal;
 489         }
 490
 491         # These take a string and run the normalization on them, without
 492         # checking for validity or any optimization etc. Input must be
 493         # VALID UTF-8!
 494         /**
 495          * @param string $string
 496          * @return string
 497          * @private
 498          * @static
 499          */
 500         static function NFC( $string ) {
 501                 return UtfNormal::fastCompose( UtfNormal::NFD( $string ) );
 502         }
 503
 504         /**
 505          * @param string $string
 506          * @return string
 507          * @private
 508          * @static
 509          */
 510         static function NFD( $string ) {
 511                 UtfNormal::loadData();
 512                 global $utfCanonicalDecomp;
 513                 return UtfNormal::fastCombiningSort(
 514                         UtfNormal::fastDecompose( $string, $utfCanonicalDecomp ) );
 515         }
 516
 517         /**
 518          * @param string $string
 519          * @return string
 520          * @private
 521          * @static
 522          */
 523         static function NFKC( $string ) {
 524                 return UtfNormal::fastCompose( UtfNormal::NFKD( $string ) );
 525         }
 526
 527         /**
 528          * @param string $string
 529          * @return string
 530          * @private
 531          * @static
 532          */
 533         static function NFKD( $string ) {
 534                 global $utfCompatibilityDecomp;
 535                 if( !isset( $utfCompatibilityDecomp ) ) {
 536                         require_once( 'UtfNormalDataK.inc' );
 537                 }
 538                 return UtfNormal::fastCombiningSort(
 539                         UtfNormal::fastDecompose( $string, $utfCompatibilityDecomp ) );
 540         }
 541
 542
 543         /**
 544          * Perform decomposition of a UTF-8 string into either D or KD form
 545          * (depending on which decomposition map is passed to us).
 546          * Input is assumed to be *valid* UTF-8. Invalid code will break.
 547          * @private
 548          * @param string $string Valid UTF-8 string
 549          * @param array $map hash of expanded decomposition map
 550          * @return string a UTF-8 string decomposed, not yet normalized (needs sorting)
 551          * @static
 552          */
 553         static function fastDecompose( $string, $map ) {
 554                 UtfNormal::loadData();
 555                 $len = strlen( $string );
 556                 $out = '';
 557                 for( $i = 0; $i < $len; $i++ ) {
 558                         $c = $string{$i};
 559                         $n = ord( $c );
 560                         if( $n < 0x80 ) {
 561                                 # ASCII chars never decompose
 562                                 # THEY ARE IMMORTAL
 563                                 $out .= $c;
 564                                 continue;
 565                         } elseif( $n >= 0xf0 ) {
 566                                 $c = substr( $string, $i, 4 );
 567                                 $i += 3;
 568                         } elseif( $n >= 0xe0 ) {
 569                                 $c = substr( $string, $i, 3 );
 570                                 $i += 2;
 571                         } elseif( $n >= 0xc0 ) {
 572                                 $c = substr( $string, $i, 2 );
 573                                 $i++;
 574                         }
 575                         if( isset( $map[$c] ) ) {
 576                                 $out .= $map[$c];
 577                                 continue;
 578                         } else {
 579                                 if( $c >= UTF8_HANGUL_FIRST && $c <= UTF8_HANGUL_LAST ) {
 580                                         # Decompose a hangul syllable into jamo;
 581                                         # hardcoded for three-byte UTF-8 sequence.
 582                                         # A lookup table would be slightly faster,
 583                                         # but adds a lot of memory & disk needs.
 584                                         #
 585                                         $index = ( (ord( $c{0} ) & 0x0f) << 12
 586                                                  | (ord( $c{1} ) & 0x3f) <<  6
 587                                                  | (ord( $c{2} ) & 0x3f) )
 588                                                - UNICODE_HANGUL_FIRST;
 589                                         $l = intval( $index / UNICODE_HANGUL_NCOUNT );
 590                                         $v = intval( ($index % UNICODE_HANGUL_NCOUNT) / UNICODE_HANGUL_TCOUNT);
 591                                         $t = $index % UNICODE_HANGUL_TCOUNT;
 592                                         $out .= "\xe1\x84" . chr( 0x80 + $l ) . "\xe1\x85" . chr( 0xa1 + $v );
 593                                         if( $t >= 25 ) {
 594                                                 $out .= "\xe1\x87" . chr( 0x80 + $t - 25 );
 595                                         } elseif( $t ) {
 596                                                 $out .= "\xe1\x86" . chr( 0xa7 + $t );
 597                                         }
 598                                         continue;
 599                                 }
 600                         }
 601                         $out .= $c;
 602                 }
 603                 return $out;
 604         }
 605
 606         /**
 607          * Sorts combining characters into canonical order. This is the
 608          * final step in creating decomposed normal forms D and KD.
 609          * @private
 610          * @param string $string a valid, decomposed UTF-8 string. Input is not validated.
 611          * @return string a UTF-8 string with combining characters sorted in canonical order
 612          * @static
 613          */
 614         static function fastCombiningSort( $string ) {
 615                 UtfNormal::loadData();
 616                 global $utfCombiningClass;
 617                 $len = strlen( $string );
 618                 $out = '';
 619                 $combiners = array();
 620                 $lastClass = -1;
 621                 for( $i = 0; $i < $len; $i++ ) {
 622                         $c = $string{$i};
 623                         $n = ord( $c );
 624                         if( $n >= 0x80 ) {
 625                                 if( $n >= 0xf0 ) {
 626                                         $c = substr( $string, $i, 4 );
 627                                         $i += 3;
 628                                 } elseif( $n >= 0xe0 ) {
 629                                         $c = substr( $string, $i, 3 );
 630                                         $i += 2;
 631                                 } elseif( $n >= 0xc0 ) {
 632                                         $c = substr( $string, $i, 2 );
 633                                         $i++;
 634                                 }
 635                                 if( isset( $utfCombiningClass[$c] ) ) {
 636                                         $lastClass = $utfCombiningClass[$c];
 637                                         @$combiners[$lastClass] .= $c;
 638                                         continue;
 639                                 }
 640                         }
 641                         if( $lastClass ) {
 642                                 ksort( $combiners );
 643                                 $out .= implode( '', $combiners );
 644                                 $combiners = array();
 645                         }
 646                         $out .= $c;
 647                         $lastClass = 0;
 648                 }
 649                 if( $lastClass ) {
 650                         ksort( $combiners );
 651                         $out .= implode( '', $combiners );
 652                 }
 653                 return $out;
 654         }
 655
 656         /**
 657          * Produces canonically composed sequences, i.e. normal form C or KC.
 658          *
 659          * @private
 660          * @param string $string a valid UTF-8 string in sorted normal form D or KD. Input is not validated.
 661          * @return string a UTF-8 string with canonical precomposed characters used where possible
 662          * @static
 663          */
 664         static function fastCompose( $string ) {
 665                 UtfNormal::loadData();
 666                 global $utfCanonicalComp, $utfCombiningClass;
 667                 $len = strlen( $string );
 668                 $out = '';
 669                 $lastClass = -1;
 670                 $lastHangul = 0;
 671                 $startChar = '';
 672                 $combining = '';
 673                 $x1 = ord(substr(UTF8_HANGUL_VBASE,0,1));
 674                 $x2 = ord(substr(UTF8_HANGUL_TEND,0,1));
 675                 for( $i = 0; $i < $len; $i++ ) {
 676                         $c = $string{$i};
 677                         $n = ord( $c );
 678                         if( $n < 0x80 ) {
 679                                 # No combining characters here...
 680                                 $out .= $startChar;
 681                                 $out .= $combining;
 682                                 $startChar = $c;
 683                                 $combining = '';
 684                                 $lastClass = 0;
 685                                 continue;
 686                         } elseif( $n >= 0xf0 ) {
 687                                 $c = substr( $string, $i, 4 );
 688                                 $i += 3;
 689                         } elseif( $n >= 0xe0 ) {
 690                                 $c = substr( $string, $i, 3 );
 691                                 $i += 2;
 692                         } elseif( $n >= 0xc0 ) {
 693                                 $c = substr( $string, $i, 2 );
 694                                 $i++;
 695                         }
 696                         $pair = $startChar . $c;
 697                         if( $n > 0x80 ) {
 698                                 if( isset( $utfCombiningClass[$c] ) ) {
 699                                         # A combining char; see what we can do with it
 700                                         $class = $utfCombiningClass[$c];
 701                                         if( !empty( $startChar ) &&
 702                                                 $lastClass < $class &&
 703                                                 $class > 0 &&
 704                                                 isset( $utfCanonicalComp[$pair] ) ) {
 705                                                 $startChar = $utfCanonicalComp[$pair];
 706                                                 $class = 0;
 707                                         } else {
 708                                                 $combining .= $c;
 709                                         }
 710                                         $lastClass = $class;
 711                                         $lastHangul = 0;
 712                                         continue;
 713                                 }
 714                         }
 715                         # New start char
 716                         if( $lastClass == 0 ) {
 717                                 if( isset( $utfCanonicalComp[$pair] ) ) {
 718                                         $startChar = $utfCanonicalComp[$pair];
 719                                         $lastHangul = 0;
 720                                         continue;
 721                                 }
 722                                 if( $n >= $x1 && $n <= $x2 ) {
 723                                         # WARNING: Hangul code is painfully slow.
 724                                         # I apologize for this ugly, ugly code; however
 725                                         # performance is even more teh suck if we call
 726                                         # out to nice clean functions. Lookup tables are
 727                                         # marginally faster, but require a lot of space.
 728                                         #
 729                                         if( $c >= UTF8_HANGUL_VBASE &&
 730                                                 $c <= UTF8_HANGUL_VEND &&
 731                                                 $startChar >= UTF8_HANGUL_LBASE &&
 732                                                 $startChar <= UTF8_HANGUL_LEND ) {
 733                                                 #
 734                                                 #$lIndex = utf8ToCodepoint( $startChar ) - UNICODE_HANGUL_LBASE;
 735                                                 #$vIndex = utf8ToCodepoint( $c ) - UNICODE_HANGUL_VBASE;
 736                                                 $lIndex = ord( $startChar{2} ) - 0x80;
 737                                                 $vIndex = ord( $c{2}         ) - 0xa1;
 738
 739                                                 $hangulPoint = UNICODE_HANGUL_FIRST +
 740                                                         UNICODE_HANGUL_TCOUNT *
 741                                                         (UNICODE_HANGUL_VCOUNT * $lIndex + $vIndex);
 742
 743                                                 # Hardcode the limited-range UTF-8 conversion:
 744                                                 $startChar = chr( $hangulPoint >> 12 & 0x0f | 0xe0 ) .
 745                                                                          chr( $hangulPoint >>  6 & 0x3f | 0x80 ) .
 746                                                                          chr( $hangulPoint       & 0x3f | 0x80 );
 747                                                 $lastHangul = 0;
 748                                                 continue;
 749                                         } elseif( $c >= UTF8_HANGUL_TBASE &&
 750                                                           $c <= UTF8_HANGUL_TEND &&
 751                                                           $startChar >= UTF8_HANGUL_FIRST &&
 752                                                           $startChar <= UTF8_HANGUL_LAST &&
 753                                                           !$lastHangul ) {
 754                                                 # $tIndex = utf8ToCodepoint( $c ) - UNICODE_HANGUL_TBASE;
 755                                                 $tIndex = ord( $c{2} ) - 0xa7;
 756                                                 if( $tIndex < 0 ) $tIndex = ord( $c{2} ) - 0x80 + (0x11c0 - 0x11a7);
 757
 758                                                 # Increment the code point by $tIndex, without
 759                                                 # the function overhead of decoding and recoding UTF-8
 760                                                 #
 761                                                 $tail = ord( $startChar{2} ) + $tIndex;
 762                                                 if( $tail > 0xbf ) {
 763                                                         $tail -= 0x40;
 764                                                         $mid = ord( $startChar{1} ) + 1;
 765                                                         if( $mid > 0xbf ) {
 766                                                                 $startChar{0} = chr( ord( $startChar{0} ) + 1 );
 767                                                                 $mid -= 0x40;
 768                                                         }
 769                                                         $startChar{1} = chr( $mid );
 770                                                 }
 771                                                 $startChar{2} = chr( $tail );
 772
 773                                                 # If there's another jamo char after this, *don't* try to merge it.
 774                                                 $lastHangul = 1;
 775                                                 continue;
 776                                         }
 777                                 }
 778                         }
 779                         $out .= $startChar;
 780                         $out .= $combining;
 781                         $startChar = $c;
 782                         $combining = '';
 783                         $lastClass = 0;
 784                         $lastHangul = 0;
 785                 }
 786                 $out .= $startChar . $combining;
 787                 return $out;
 788         }
 789
 790         /**
 791          * This is just used for the benchmark, comparing how long it takes to
 792          * interate through a string without really doing anything of substance.
 793          * @param string $string
 794          * @return string
 795          * @static
 796          */
 797         static function placebo( $string ) {
 798                 $len = strlen( $string );
 799                 $out = '';
 800                 for( $i = 0; $i < $len; $i++ ) {
 801                         $out .= $string{$i};
 802                 }
 803                 return $out;
 804         }
 805 }
 806
 807 ?>